Plasmonic Substrates Do Not Promote Vibrational Energy Transfer at

Letter pubs.acs.org/JPCL

Cite This: J. Phys. Chem. Lett. 2018, 9, 49−56

Plasmonic Substrates Do Not Promote Vibrational Energy Transfer at Solid−Liquid Interfaces Jan Philip Kraack,* Laurent Sévery, S. David Tilley, and Peter Hamm* Department of Chemistry, University of Zurich, Winterthurerstrasse 190, CH-8057 Zurich, Switzerland

J. Phys. Chem. Lett. 2018.9:49-56. Downloaded from pubs.acs.org by IOWA STATE UNIV on 01/19/19. For personal use only.

S Supporting Information *

ABSTRACT: Intermolecular vibrational energy transfer in monolayers of isotopically mixed rhenium carbonyl complexes at solid−liquid interfaces is investigated with the help of ultrafast 2D Attenuated Total Reflectance Infrared (2D ATR IR) spectroscopy in dependence of plasmonic surface enhancement effects. Dielectric and plasmonic materials are used to demonstrate that plasmonic effects have no impact on the vibrational energy transfer rate in a regime of moderate IR surface enhancement (enhancement factors up to ca. 30). This result can be explained with the common image-dipole picture. The vibrational energy transfer rate thus can be used as a direct observable to determine intermolecular distances on surfaces, regardless of their plasmonic properties.

I

nteractions of molecules at surfaces are extremely important in different fields of current research, such as heterogeneous catalysis, artificial light-harvesting or molecular sensing and recognition. However, it is challenging to directly observe the dynamics of interactions spectroscopically. This is particularly true for molecules that do not form chemical bonds, through which mutual interactions could occur, but rather interact “through space” in an electrostatic way, e.g., by dipole−dipole coupling. One of the most prominent ways to witness such interactions is energy transfer between different chemical entities.1−4 This effect is well established in the field of electronic spectroscopy with visible light (Förster energy transfer), as well as for NMR spectroscopy (NOESY), where energy transfer has been observed and used to elucidate structural properties of various samples.1−3,5,6 For vibrational spectroscopy in the mid-infrared (IR) range, intermolecular vibrational energy transfer through space is, however, much less frequently observed, but still holds great promise to facilitate structural investigations of samples, such as relative molecular distances and orientations.4,7−15 It is therefore strongly desirable to develop experimental strategies, through which vibrational energy transfer can be engineered. The vibrational energy transfer rate kET is commonly described by a mechanism based on the transition dipole coupling VDD (eq 1), whose dominant factors are the mutual distance of a donor and an acceptor r, the contributing transition dipole moments μD and μA of donor and acceptor, respectively (which we will assume to be the same μD = μA = μ), and a geometry factor κ describing the orientation of the transition dipoles relative to a vector connecting the two molecules. All factors enter with a strongly nonlinear dependence into the vibrational energy transfer rate kET:1−4,11,13 © 2017 American Chemical Society

2 kET ∝ VDD = κ2

μD2 μA2 r

6

= κ2

μ4 r6

(1)

in close analogy to NOESY in NMR,5 or Förster energy transfer in fluorescence spectroscopy.6 In order to render the intermolecular distance small, immobilization of molecules at surfaces is a beneficial approach, since packing densities in molecular monolayers are generally high with subnanometer intermolecular distances.16,17 Thus, molecular monolayers are prime candidates to observe vibrational energy transfer, and quite some experimental effort went into this direction recently.8,18−25 We have shown in ref 8 that in addition to a short distance, a large transition dipole μ is required to compete with the generally short lifetime of vibrational transitions. As of today, intermolecular vibrational energy transfer for a molecular monolayer has been observed only for a metal−carbonyl complex, owing to its exceptionally large transition dipole in combination with a comparably long lifetime.8 In essence, the equivalent of a “Förster radius” for vibrational transitions (i.e., the distance at which the vibrational energy transfer rate equals the inverse lifetime) is small (∼3−4 Å) even for metal− carbonyls that exhibit a very high transition dipole already. This small value limits the versatility of the approach, and even higher transition dipoles would be desirable.26−28 In that regard, it is well-known that the IR absorption cross section of molecules on metallic surfaces are often strongly enhanced by plasmonic effects, either on rough metal surface or even more so with specifically designed nanostructures.29−32 The absorption cross section A of a vibrational transition scales as Received: October 27, 2017 Accepted: December 13, 2017 Published: December 13, 2017 49

DOI: 10.1021/acs.jpclett.7b02855 J. Phys. Chem. Lett. 2018, 9, 49−56

Letter

The Journal of Physical Chemistry Letters A ∝ μeff2

the bipyridine ligands via different linkers (“L”, Figure 1), depending on the type of the surface. That is, two carboxyl groups at the bipyridine ligands are used for immobilization in case of oxide surfaces (TiO2 and ITO), whereas two methyl thioether groups are employed in case of a metal surface (Au). In the latter case, we use thioethers rather than thiols to prevent dimerization of the molecules in solution by oxidative formation of disulfide bonds. The employed Re carbonyl molecules exhibit several prominent benefits regarding their use as reporters of vibrational energy transfer. First, the transition dipole of the totally symmetric carbonyl-stretching vibration is large, as evidenced by the high absorption coefficient of ≈3400 M−1 cm−1.8 This aspect facilitates a reasonable absorption from only monolayer thin samples at the interface (eq 2) as well as a reasonably fast vibrational energy transfer rate (eq 1). Second, the molecules form stable bonds to the surface due to their two easily accessible functional groups for immobilization and thus feature a very low degree of structural flexibility. Third, the molecules exhibit a considerable time scale for vibrational relaxation with lifetimes larger than 20 ps, which allows a long temporal observation window (about 60 ps). Fourth, the totally symmetric A′(1) stretch vibration of the carbonyl ligands is located in an experimentally well accessible spectral region around 2000 cm−1 and modes from different molecules can easily be made spectroscopically distinguishable by use of 13C isotope substitution of the carbonyl groups.8 That is, the A′(1) vibration of Re(12CO)3Cl absorbs at about 2025 cm−1, whereas that of Re(13CO)3Br is located at about 1976 cm−1. We have recently shown that vibrational energy transfer can be observed between these two modes when the sample is immobilized on ITO.8 In the present study, the same molecular system has been immobilized on a series of different substrate layers, representing different levels of plasmonic properties. As the one extreme case, TiO2 is a purely dielectric material, and the incident IR light interacts with the adsorbate molecules only, but cannot polarize the TiO2 layer. Indeed we measured an “enhancement factor” of EF = 1, i.e., no enhancement (Figure SI1e,f) for TiO2, which thus acts exclusively as a layer for sample immobilization at the interface. Conversely, metal layers represent the other extreme case, since they are strongly polarizable.29,31,32,47 Particularly, Au layers of up to a few nanometre thickness have been shown previously to yield a significant surface enhancement in 2D ATR IR spectroscopy, which stems from a polarization of the metal substrate by the incident IR light along with possible other enhancement mechanisms such as chemical contributions.48,49 For our specific sample preparation, we measured an enhancement factor of about EF = 30 (Figure SI1a,b). Between the two extreme cases, indium−tin-oxide (ITO) is chosen as a material exhibiting electrical conductivity50 and a high mobility of electrons, which has been reported to facilitate weak plasmonic properties.51−53 For our particular sample preparation, we have measured an enhancement factor of about EF = 2 (Figure SI1c,d). Besides orientational effects by aligning the molecules on the surface,35,48,54 this small enhancement factor ITO might indeed reflect a weak plasmonic signal enhancement; nevertheless ITO is certainly more on the side of a dielectric than on the side of a metal. Comparing the dynamics of vibrational energy transfer on these surfaces with otherwise as identical as possible sample conditions allows us to study the impact of substrate polarizability and plasmonic surface enhancement on the vibrational energy transfer mechanism.

(2)

where μeff is an effective, potentially enhanced transition dipole. Considering eqs 1 and 2, one might therefore assume that vibrational energy transfer is enhanced as well on plasmonic surfaces, provided that the transition dipole moments relevant for both effects are the same. The aim of this paper is to test whether that is indeed the case. To that end, we employ a surface-sensitive variant of 2D IR spectroscopy,33−35 since this allows direct observation of vibrational energy transfer through the identification of cross peaks in the signals. Specifically, we use 2D ATR IR spectroscopy, which has been introduced recently as a new and very flexible technique for surface vibrational characterization of adsorbates in different environments.35−37 The information content in 2D ATR IR spectra is analogous to that of “classic” 2D IR spectroscopy in transmission geometry,33,34 but rather exploits evanescent waves at the interface of an ATR element to excite and probe the sample. The molecule under study is a metal−carbonyl complex, Re(4,4′-dilinker-2,2′-bipyridine)(CO)3Cl (Re(CO)3Cl; see Figure 1). The same24,25,38−41 or very similar metal carbonyl

Figure 1. Sample system as used in this study. Sample molecules (Re(4,4′-dilinker-2,2′-bipyridine)(13CO)3Br (Re(13CO)3Br) and Re(4,4′-dilinker-2,2′-bipyridine)(12CO)3Cl (Re(12CO)3Cl) are covalently adsorbed as about 1:1 mixtures on titanium dioxide (TiO2), indium− tin-oxide (ITO), and gold (Au) surfaces. Depending on the type of surface, the linker L is a carboxyl group for oxide surfaces (TiO2 and ITO), or a methyl thioether (S-Me) for Au.

complexes18,42−46 on different types of surfaces have already been investigated quite intensively recently with the help of 2D IR and 2D-SFG spectroscopy. The importance of surface heterogeneity,24,25,38,39,41,44 the intramolecular dynamics within the immobilized molecule,40 as well as the coupling to the surface42,43 have been pointed out. Moreover, vibrational couplings have been observed between different sub-bands when the complexes aggregate on the surface.24,25 Form these works, it is well established that the adsorbate dynamics strongly depend on subtle properties of the substrate. A detailed sketch of the complete sample system under investigation in the present study is shown in Figure 1. Mixed (1:1) monolayers with Re(4,4′-dilinker-2,2′-bipyridine)( 13 CO) 3 Br (Re( 13 CO) 3 Br) and Re(4,4′-dilinker-2,2′bipyridine)(12CO)3Cl (Re(12CO)3Cl) head groups are covalently adsorbed to different types of surfaces, which are deposited on the reflecting plane of a CaF2 ATR prism. The molecules are attached to the surfaces at the 4,4′-positions of 50

DOI: 10.1021/acs.jpclett.7b02855 J. Phys. Chem. Lett. 2018, 9, 49−56

Letter

The Journal of Physical Chemistry Letters

similar for the TiO2/ITO samples and slightly smaller for the Au layer. To elucidate the detailed kinetics of vibrational energy transfer, we measured 2D ATR IR spectra at a full series of population delays that covers multiples of the vibrational lifetime of the complexes. Figure 3a−c shows the temporal

Figure 2 shows 2D ATR IR spectra of the mixed monolayer sample system Re(13CO)3Br/Re(12CO)3Cl for two representa-

Figure 2. 2D ATR IR spectra from the A′(1) carbonyl stretching region at different waiting times (0.5 and 21 ps) for the different sample systems shown in Figure 1 and immersed in MeOH. Samples adsorbed to (a,b) TiO2, (c,d) ITO, and (e,f) Au surfaces are compared. The circles in panel a indicate the different contributions for the two Rhenium complexes. Blue and red signals represent ground state bleach and excited state absorption signals, respectively. Cross peaks appear with increasing population delays, indicating vibrational energy transfer between the molecules. Signals are shown normalized to maximum ground state bleach intensity at each population delay to facilitate comparison.

Figure 3. (a−c) Kinetic traces of ground state bleach signals of diagonal (circles) and cross peaks (triangles) together with kinetic fits (red lines) using the model from (d). Note that the cross peak signals are magnified by a factor of 10. (d) Schematic representation of vibrational energy transfer pathways and vibrational relaxation in the investigated sample systems. ET = energy transfer, bw = backward, fw = forward.

tive population delays (i.e., 0.5 and 21 ps) for the three different surface materials, i.e., Figure 2a,b for TiO2, c,d for ITO layers, and e,f for Au. All samples have been incubated with methanol during the experiments. The measurements for ITO are in principle analogous to ref 8, but have been remeasured for consistency reasons. For all samples, the 2D ATR IR spectra at early population delays (0.5 ps) exhibit ground state bleach (blue) and excited state absorption signals (red) on the diagonal line. The signals are strongly elongated along the diagonal, reflecting the surface heterogeneity and a distribution of local solvation environments of the adsorbates.35 As the population time is increased to a few tens of picoseconds, weak cross peaks start to appear for each sample system. These cross peaks originate from vibrational energy transfer between the different molecules at the interface.8 Interestingly, the relative intensity between the cross peaks and the diagonal peaks is very

evolution of both diagonal (circles) as well as cross peaks (triangles) up to a population delay of 60 ps. For all samples, the diagonal peaks decay monotonically, whereas the cross peak signals initially increase, peak at population delays between 13 and 25 ps, and then decay again. We observe similar maximum cross peak intensity and similar delays for the appearance of the cross peak maxima for TiO2 and ITO layers (Figure 3a,b), whereas the intensity of the cross peaks is lower by approximately a factor of 2 in the case of the Au layers, and the cross peak maximum is significantly delayed (Figure 3c). The data are fitted (red lines) with the kinetic model depicted in Figure 3d.8 In that scheme, both molecules can be excited 51

DOI: 10.1021/acs.jpclett.7b02855 J. Phys. Chem. Lett. 2018, 9, 49−56

Letter

The Journal of Physical Chemistry Letters

Note that this is a huge discrepancy, which would safely overcompensate modest errors in the estimate of the intermolecular distance resulting from, e.g., an inaccurate measurement in the surface coverage or the formation of domains on the surface. The same holds for variations of the orientational factor κ in eq 1, which might be smaller on Au, since the molecules are oriented slightly differently due to the different geometry of the methyl-thioether linker. Traditionally, the signal enhancement in surface spectroscopy is separated into an “electromagnetic” and a “chemical” contribution.47,48,57−63 The chemical enhancement mechanism takes into account substrate−adsorbate electronic interactions, resulting, for instance, in delocalized electronic states, chargetransfer states between the surface and the adsorbate, or change of electronic structure within the molecule. In essence, a chemical enhancement is the result of different partial charges within the molecule, which should enter the transition dipoles in eqs 1 and 2 in the same way. The absent speed-up of vibrational energy transfer on Au strongly suggests that such a chemical enhancement mechanism is not a very prominent effect here. That might indeed be expected, given the linker between the Re(CO)3-group and the surface, which decouples them electronically from each other. Conversely, the electromagnetic mechanism61−66 originates from the polarizability of the metallic substrate. We start with noting that our Au surfaces are nanostructured (Figure SI 2) with a plasmon resonance at about 625 nm (Figure SI 3), while we investigate surface enhancements of vibrational transitions that are very far off-resonant in the mid-IR.48 We can therefore consider the electric field of the IR light as quasi-static, simplifying the discussion significantly.54 Figure 4a illustrates

independently (stars) and can afterward either transfer the excess vibrational energy to nonexcited neighboring molecules with time constant τET, or undergo vibrational relaxation with time constants τ1,2. For all three systems, we observe that vibrational relaxation occurs in a biexponential manner, with time constants of 2−3 ps (fast) and 20−21 ps (slow) and amplitude ratios between these two contributions of about 0.3/ 0.7. This biexponential decay is in qualitative agreement with previous observations and has been attributed to intramolecular vibrational energy redistribution (IVR) between /the symmetric and asymmetric stretching modes of the carbonyl ligands,8,40,55,56 but surface heterogeneity might contribute as well. Regarding the cross peak dynamics, we retrieve time constants for the vibrational energy transfer from 13CO to 12 CO (upper cross peak in Figure 2) of about 90 ps on TiO2, 110 ps on ITO (which within error agrees with the result of ref 8 90 ps), and 260 ps on Au. Interestingly, the vibrational energy transfer on the plasmonic Au layers is even slower compared to that on dielectric substrates, despite the plasmonic enhancement of the absorption cross section. Evaluating eq 1, we have shown in our previous work that our preparation conditions result essentially in a closest possible packing of the molecules on ITO with intermolecular distances of about 4−5 Å.8 In order to compare the results on the different substrates, we need to estimate the relative surface coverages, since the intermolecular distance strongly affects the transfer dynamics (eq 1). We do this comparison here for Au and ITO, since the preparation of both materials by sputtercoating leads to the formation of nanoparticle layers with similar dimensions (see Figure SI 2). Contrasting to that, TiO2 has been prepared by atomic-layer-deposition (ALD), likely revealing a more uniform layer. To determine the relative surface coverages on Au and ITO, we use the in situ measured IR absorbance in conjunction with the experimentally determined enhancement factors.48,54 That is, the ATR IR absorbance for the Re carbonyl complexes on Au and ITO have been determined to about 36 mOD (Figure SI 1a) and 3.5 mOD (Figure SI 1c)), while the corresponding enhancement factors are 30 and 2, respectively (Figure SI 1a−d). As the IR absorbance (μ2) scales linearly with the enhancement factor EF, we obtain a surface coverage c on Au that is about 70% of that on ITO. Equation 1 predicts an r−6 ∝ c−3 scaling of the vibrational energy transfer rate, when assuming randomly distributed molecules on the surface without the formation of any domains. Based on the larger intermolecular distance, one would expect a three times slower transfer rate for the complexes on the Au layers as compared to ITO layers, in rough agreement with the experimental results (Figure 3). This larger distance is indeed expected for instance by considering that the binding motif of the S−Me linkers for the complexes on Au will be slightly different as compared to the COOH linkers at the bipyridyl ligand in the case of oxide surfaces. Moreover, the increased sterical hindrance imposed by the methyl group is likely to result in a larger intermolecular spacing between the complexes. This result, in turn, implies that plasmonic effects do not enhance the vibrational energy transfer rate. That is, if the signal enhancement measured for the Au sample, which scales as EF ∝ A ∝ μ2eff (eq 2), would contribute to the transition dipole coupling as well, one would expect an EF2 ≈ 230-fold acceleration of the vibrational energy transfer rate relative to that on ITO (eq 1). This is clearly not observed; to the contrary, the vibrational energy rate goes down by a factor 2−3.

Figure 4. (a) Two interacting dipoles μ⃗ 1and μ⃗ 2 atop of a metal surface, which we assume to be perpendicular to the surface, and which induce image dipoles μ⃗ i1 and μ⃗ i2. (b) In terms of interaction energies, this configuration can be reduced to one without metal and with only three (eff) (eff) remaining dipole−dipole interaction terms V(eff) 1,2 , V1,2i , and V1,1i (see SI for the derivation). The separation of the two dipoles from the surface is R, and that among each other r.

two interacting dipoles μ⃗ 1and μ⃗ 2 atop of a planar metal surface as the simplest possible geometry, together with their corresponding image-dipoles μ⃗ i1 and μ⃗ i2. For the absorption cross section of each one of these transition dipoles, both dipole and corresponding image dipole add up constructively, μeff = μ + μi = 2μ, which, according to eq 2 reveals an enhancement factor of EF = 4. With regard to the transition dipole coupling (eq 1), on the other hand, one can show that the configuration of Figure 4a can be reduced to that of Figure 4b without any metal and only two terms contributing to the coupling, namely, the normal 52

DOI: 10.1021/acs.jpclett.7b02855 J. Phys. Chem. Lett. 2018, 9, 49−56

Letter

The Journal of Physical Chemistry Letters

absorption cross section is significantly enhanced. In other words, the effective transition dipoles entering eqs 1 and 2 are not the same. By comparing vibrational energy transfer between the symmetric carbonyl stretching vibrations on three different substrates, we find that the rate is even slightly slower on plasmonic metal layers as compared to purely dielectric layers, while they are about the same on the oxides TiO2 and ITO. The transition dipole and its corresponding image dipole add up constructively for the absorption cross section (eq 2) on a length scale that is determined by the roughness of the surface. On the contrary, the very small “Förster” radius for vibrational energy transfer in essence suppresses the dipole coupling to the image dipole for the purpose of vibrational energy transfer (eq 1). This result, in turn, allows one to use the vibrational energy transfer rate as a measure to determine intermolecular distances on a surface, regardless of its plasmonic properties. Both linker and surface chemistry are about the same in our TiO2 and ITO preparations, and consequently also the vibrational energy transfer rates, while vibrational energy transfer is somewhat slower on gold, presumably due to the different structures of the methyl-thioether linker group. Similar experiments on purposely designed plasmonic nanostructures, which are resonant with the IR transition and consequently reveal significantly larger enhancement factors, will be most interesting to see whether the conclusions drawn here can be generalized to those structures.

dipole−dipole interaction terms between the two transition dipoles V(eff) 1,2 as well as to the corresponding mirror transition dipole V(eff) 1,2i , while somewhat nonintuitively, the coupling between the two image transition dipoles (Vi1,i2) is lumped into V(eff) 1,2 and does not appear a second time (see SI for a detailed derivation). In simple words, this results from the fact that the electric field inside the metal vanishes, so μ⃗ i1 cannot interact with μ⃗i2. In our specific case, the distance between two dipoles exhibiting vibrational energy transfer is r = 4−5 Å,8 while the distance of the CO groups to the surface is R ≈ 10 Å, hence the distance to the image dipole is 2R ≈ 20 Å, about 4 times larger than the intermolecular distance. Since the transition dipole coupling scales as r−3, we may safely neglect V(eff) 1,2i and find that the transition dipole coupling is determined by V(eff) 1,2 only, as if the metal surface would not be present. That derivation explains why we do not observe any acceleration of the vibrational energy transfer rate despite the fact that we do observe an enhancement of the absorption. Note that Figure 4 discusses the situation on a perfectly planar surface, in which case an enhancement factor of EF = 4 is expected for absorption. That enhancement factor is obtained as long as the dipole approximation holds for the total dipole μeff, i.e., whenever the distance of the dipole from its image dipole 2R is much smaller than the wavelength of the light (which we can indeed safely assume). The enhancement factor is larger in our specific experiment (EF ≈ 30) due to the roughness of the surface and possibly hotspots between Au nanoparticles on the surface.48 We expect that the argument of Figure 4 remains in essence the same; however, in that case it is the roughness of the surface, as characterized by the typical curvature radius, that determines the distance from the surface below which the enhancement factor is large.54 The IR enhancement from the Au layers, which we have used here, are rather small (